[gfortran,patch] Fix code generated for NINT

[Janne Blomqvist]

On 8/1/07, FX Coudert <fxcoudert@gmail.com> wrote:
> Yes, I know. I have lots of code myself which will probably get a
> large hit in performance due to that patch, and I was really
> reluctant to do it for such a simple case, but I can't find another
> way of fixing it :(

Does builtin_lround expand inline to the previous

INT(x + ((x > 0) ? 0.5 : -0.5))

if -fast-math is enabled? If not, I suppose a PR is warranted?

-- 
Janne Blomqvist